Managing Security Groups and Assignments
PlaidCloud’s security and access management is straightforward. A member is granted or denied access based on the groups in which a member is associated. Adding or changing a member’s security association is easily customizable.
Managing Security Groups
Section titled “Managing Security Groups”Security groups can be added, updated, or deleted.
To manage security groups:
- Open Identity
- Select Security groups in the rail (this will display a list of existing groups)
- To add a group, click New group
- To edit permissions of a group, click on the left-most icon
To manage group members:
- Open Identity
- Select Security groups in the rail
- Click the Member icon
- Tick the members who should belong to the group, and clear the ones who should not. Use the filter box to find people. See Assigning With the Picker
- Click Update Assignments
Assigning With the Picker
Section titled “Assigning With the Picker”Wherever you give people, groups or other items access to something, one checklist replaces the old pair of lists. Every row has a checkbox, and a ticked row is assigned. Each row shows the name and the email in separate columns. You use it to set a group’s members, a member’s security groups, an account’s members, groups, agents and owners, a connection’s access, a distribution list’s workspace members, and the workflows in a Weave readiness set.
- Filter. Type in the filter box to narrow the list. Matching is partial and ignores case, so
smithfinds every name or email containing it, and@example.comfinds everyone at that domain. Typing several words keeps only rows that match all of them. - Paste a list of emails. Paste several email addresses, separated by spaces, commas, semicolons or new lines, and the list shows exactly those people. You can mix in whole domains such as
@example.com. A line under the filter says how many pasted entries matched and lists any that were not found. Assign followed by the number matched ticks everyone who matched in one click. - All, Assigned and Not assigned. The three buttons above the list switch between everything, only what is assigned, and only what is not, each with a count. They combine with the filter.
- Assign all shown and Unassign all shown. These tick or clear every row the filter and view currently show, and leave hidden rows alone. A line above the list says how many rows are shown out of the total.
- Will add and Will remove. Nothing is saved until you click the save button. A row you have ticked is tinted and shows a Will add pill, and a row you have cleared shows Will remove. The summary above the list shows how many are assigned and the pending changes. The summary reads, for example,
12 assigned · +3 to add · −1 to remove. Undo changes puts everything back to how it was when you opened the window. - Save only when something changed. The save button, such as Update Assignments, Update Members or Update Groups, is available only when you have made a change.
- Keyboard. Space ticks or clears the selected row. Enter in the filter box does not save, so pressing it while searching can never change access.
When you can see who has access but cannot change it, for example on a connection you do not own, the picker is read-only: a note says so, and it lists only the people already assigned, with their initials in place of checkboxes and no buttons that change them.
Managing Row Access
Section titled “Managing Row Access”Row Access restricts which rows of a table a security group can see, based on the value of one column. For example, a region column can restrict access, with a group granted East and Central limited to rows carrying those values. Everything for a project and column — picking the column, choosing where its values come from, and deciding which group sees what — happens in a single screen, and saving it takes effect immediately.
Setting up and changing Row Access takes a project Architect or a tenant admin. Anyone else who opens a project’s Security tab is told that row access is Architect-only, and the entry points below don’t appear for them.
Opening Row Access
Section titled “Opening Row Access”Every entry point opens the same Row Access screen:
- Table Explorer’s toolbar has a Row Access button, with a badge when the table already restricts rows. It opens on the table’s first restricted column, if it has one, and you can switch to any other column of the table.
- A column’s facet menu has Restrict rows by this column…, which opens Row Access on that column. On a column that already restricts rows, the item reads Rows restricted by this column — edit….
- A project’s Security tab lists every restricted column on the project, with the tables it applies to, the groups that hold grants, and its Sync state. Select a column to edit it, or click Restrict a column… to start a new one: pick a table, then a column. Columns that already restrict rows are marked with a lock.
- Identity → Member, on a member’s Row Access tab, and Identity → Security → Security Groups, on a group’s Row Access summary. Open Row Access (or Edit) opens the selected column; Add restricts another column in that row’s project.
Opened on a column, the screen stays on that column. Opened with Restrict a column… or Add, it lets you choose one.
The Row Access Grid
Section titled “The Row Access Grid”To restrict a table by a column:
- Open Row Access from any of the entry points above, and choose the column if you didn’t arrive with one — the picker offers the table’s real columns, not free text.
- Choose a Value Source — Column (the distinct values of a table’s column), Dimension (nodes of a dimension), or List (values you type in). The value type follows the column’s data type.
- For each security group, check the values that group may see. The grid lists the groups with access to the project, plus any group that already holds a grant on the column.
- Set Everyone else to sees all rows or sees no rows — this covers any group with no checks in the grid.
- Click Save.
Saving is live. PlaidCloud’s own queries follow the change at once and your Superset dashboards within about a minute — there’s no separate step to apply it, and Save stops to ask only when a grant holds a value its source no longer has (see below). The screen’s sync status shows how far the change has got; see Keeping Dashboards in Sync.
The grid also handles the details for you:
- Many values. Values are sorted. When a column has more values than fit, the grid says “More values than fit here — type to search.” and the filter box searches the whole column, keeping every value a group already holds.
- Values the source no longer has. A value a group was granted stays in the grid when its source loses it — no row of the table has it now, the dimension no longer has the member, or the list no longer has it. Its column header carries a warning icon, and hovering it says why. Select All, Invert and Paste leave such a value as each group holds it. Nothing is taken out of a grant unless you untick it: when you click Save while a group still holds such a value, Row Access names the groups and values in Values No Longer in the Source and asks. Keep and Save saves them as they are, Remove and Save takes them out of every group’s grant first, and closing the question saves nothing. If removing them would leave a group with nothing checked, the question says whether that group would then see all rows or none, following Everyone else. A check you haven’t saved yet goes with its value. When a column has more values than fit, or its table isn’t built yet, Row Access can’t tell which values the table lacks, so it keeps every granted value without marking any.
- Blank values. One Blank values — visible to every group with a grant checkbox decides the rows with no value in the column. Tick it and every group with any grant on the column sees those rows.
- A table that isn’t built yet shows Build this table first (run the workflow that creates it) instead of an empty list.
- Save waits for the values on screen. While values load — after you switch the value source, dimension or table, or apply a template — Save is disabled with the tooltip “Save is available once this source’s values load.” If a load fails, the grid says so and asks you to pick the source again; nothing is saved until a load succeeds.
One Column Name, Every Table
Section titled “One Column Name, Every Table”A restriction belongs to a column name on the project, not to a single table. Once region restricts rows, every table in the project with a column named region is filtered by it, in PlaidCloud’s own queries and in dashboards alike. The grid’s Applies to list shows those tables. It is read-only, because the column name decides it, and a table you add later with the same column is covered as soon as it’s published.
When no table in the project has a column with that name, the window warns that the restriction won’t filter any dashboard until a table does, in addition to showing Applies to 0 tables. Add the column to a table, or correct the column name, and the warning clears.
Stopping a Restriction
Section titled “Stopping a Restriction”Stop restricting, on the grid, removes the restriction after an inline confirmation — “Stop restricting rows by ‘region’? Every group’s grant on it is removed and everyone sees its rows.” Row Access then confirms “Rows in ‘region’ are no longer restricted.” The column’s Superset dataset goes with it, unless charts still use it; in that case Row Access says the dataset stays and how many charts use it.
A column can’t be renamed or dropped while it restricts rows. An Architect is told to stop restricting it in Row Access first; anyone else is told to ask a project Architect.
Who Can Change a Project That Restricts Rows
Section titled “Who Can Change a Project That Restricts Rows”A project’s workflows, steps and user-defined functions run as the project’s own service identity, which reads every row. Whoever can change them could make a step copy rows they aren’t granted into a table they can read, so restricting rows on a project also restricts who may change what the project does. Once a project restricts at least one column:
- Automation is Architect-only. Only project Architects and tenant admins can change workflows, steps, user-defined functions, schedules, workflow and project variables, and notify email templates — and, for the same reason, import a workflow bundle, convert an Alteryx workflow, or import, copy or restore into the project, or change its connection environment, default document account or ERP posting mode. Everyone else can still run the project’s workflows: Start, Stop, Resume and Run Selected Step(s) stay available. The step editor, the workflow’s authoring controls, the UDF editor and the schedule form open read-only for them, with a banner that names the Architects to ask: “This project restricts rows, so only its Architects can change workflows, steps, UDFs, schedules, variables and notify templates. You can still run them.”
- Dimensions a restricted column reads are Architect-only. A dimension that decides who sees which rows — the value source of a restricted column — can be changed only by a project Architect or a tenant admin, and by the project’s own workflows. That covers adding, moving, renaming, copying, clearing, loading and restoring its nodes, and the dimension itself. Anyone else is told “This dimension decides who sees which rows of ‘region’, so only a project Architect can change it.” Dimensions no restriction reads are unaffected.
- A grants source table is Architect-only. A table bound as a column’s source of grants can have its rows changed, or be deleted, only by a project Architect, a tenant admin or the project’s workflows — through the Data Editor, an import, a connector load or the API alike.
- Copying the project is Architect-only, and the copy says “This copy is not row-restricted.” — the new project carries no restriction until you set one up.
- Restoring an earlier version is a change like any other. Restoring the project’s own settings, or a dimension listed above, takes a project Architect. A restore always applies to the item the version belongs to, so it can’t be pointed at another project to get around this — see Version History.
- Editing data in a restricted table needs an Architect, as described under Row Access and the Data Editor.
- DataFrame writes, record saves and clears into a restricted table change only your own rows unless you see every row, as described under Loading Data Into a Restricted Table.
- Imports, copies, branch switches, restores and deletes of a restricted table are refused while it holds a row you cannot see, as described under Loading Data Into a Restricted Table.
Before the first column on a project is restricted, Row Access tells you who this affects: “Restricting rows here means only Architects can change this project’s workflows, steps, UDFs, schedules and notify templates. N members can change them today; they will still be able to run them.”
Runs keep working exactly as before, scheduled ones included, because they run as the project’s service identity. What a run produces follows the design its Architects gave it: an output table without the restricted column, an export file, a notification email or a log line is not filtered for whoever reads it.
Values given at the start of a run. Anyone who can run a workflow can start it with values of their own for its variables — the App Runner does this with its form’s answers, and the API and AI tools can too. Those values hold for that run only, including the workflows it starts with a Workflow: Run step: they are never saved on the workflow, and no other run, export or reader sees them. Because the run executes as the service identity, the values can change what that run reads and writes, so grant permission to run workflows to people you trust with that.
Projects that restrict no rows. These rules take effect only once a project restricts a column. On a project that restricts none, anyone who can change the project’s automation can write a step or user-defined function that — running as the project’s service identity — changes the project’s access, making themselves one of its Architects. That is by design: such a project promises no row isolation. Keep authoring rights there to people you’d trust as Architects, or restrict a column.
Keeping Dashboards in Sync
Section titled “Keeping Dashboards in Sync”Row Access shows where a change stands on the way to your dashboards, on the grid and in the Sync column of the project’s Security tab:
| Status | What it means |
|---|---|
| Waiting | A change is saved and on its way to dashboards. It normally clears within about a minute. |
| Retrying | A send failed and is being tried again. The reason for the failure is shown next to the status, with Retry to send again right away. |
| Not applied yet | The project’s restrictions have not been sent to dashboards yet. |
| Applied | The last attempt to send the restrictions to dashboards succeeded. |
| In sync with dashboards | Row Access compared what dashboards enforce with what it would write, and they match. |
| Dashboards differ | Dashboards don’t enforce what Row Access would write — a rule was changed or deleted in Superset, or a table’s published view can’t be checked. |
| Error | Sending the restrictions failed. The reason is shown next to the status. |
Once a change reaches Applied, the grid checks it against dashboards and moves on to In sync with dashboards or Dashboards differ. Dashboards differ and Error both offer Retry, which sends the project’s restrictions again. If the status can’t be read at all, the grid says “Couldn’t check sync status.”
An Error caused by one table names the table and says what to do — for example, a table published without its restricted column (“Republish or unpublish it.”), or a table with a restricted column that isn’t published, which dashboards can’t filter until it is. The status stays on Error and names the unpublished table until you publish it, or restrict the column on a published table instead. A table like that doesn’t hold up the rest of the project: every other table is kept in step.
Dashboards follow the same grants as PlaidCloud’s own queries:
- Every published table with the column is filtered, not only the table you were looking at when you set the restriction up.
- A group granted only some values is filtered even when everyone else sees all rows, on every table whose dashboard dataset exists.
- Project Architects see every row, in dashboards as they do in Table Explorer and the rest of PlaidCloud. This applies to the project’s Architect groups, and to the Architects of a project shared with individual members. Taking away Architect access takes effect in dashboards immediately, whether you change the member’s role, remove the member or their group from the project, change the project’s access type, restore an earlier project version, or delete the group.
- Security group membership changes reach dashboards on their own. Adding someone to a group or removing them in PlaidCloud, disabling or deleting a member, deleting a group, or adding a new member who joins your default security groups updates what they see in dashboards within about a minute, on every project that grants that group values, without anyone reopening Row Access. A membership change made outside PlaidCloud — directly in Keycloak, or by single sign-on placing someone in a group at sign-in — reaches dashboards within about 15 minutes, with nothing to save or retry.
Dimension Grants
Section titled “Dimension Grants”With Dimension as the value source, check a node to grant everything under it. The grant follows the dimension: a member added under a checked node is granted to that group on the next sync after the dimension is saved, and a member moved out from under it, or removed, stops being granted — nobody needs to reopen the grid. When you do reopen it, the nodes you checked show checked again.
A member you checked on its own, rather than through a node, stays granted when it’s removed from the dimension, and the grid marks it as a value the source no longer has.
Unchecking anything beneath a checked node unchecks the node, and the group keeps only the members still checked. Checking that member again doesn’t check the node again — click the node’s own box for that. Paste, Select All and Invert check members, never nodes.
A dimension member whose name can’t be used as a row-access value is refused when you save, naming it. If such a member appears later, its rows are not granted to anyone, and the sync status names it: “Dimension ‘…’ member ‘…’ cannot be used as a row access value, so its rows are not granted. Rename it in the dimension.”
Binding Grants to a Source Table
Section titled “Binding Grants to a Source Table”Instead of checking values by hand, you can point a restricted column at a source table that already holds the answer — one row per group and the value it may see. Restrict the column in the grid first, then add a Set Row Access workflow step naming the column, the source table, and the source table’s group and value columns. The step’s first run binds the column to that table: the column’s hand-set grants are cleared, and its grants are set from the table’s rows.
Once bound, the column is source-owned: the Row Access grid shows it locked, with a badge naming the source table, and hand-editing its grants is refused until you click Remove Binding.
Each later run of the step re-reads the source table and re-applies the grants — schedule it after the workflow that loads or refreshes that table, and a group added, removed, or reassigned in the source data reaches Row Access on the next run, with nobody editing the grid by hand.
Because the source table decides who sees what, only a project Architect, a tenant admin or the project’s own workflows can change its rows or delete it. Anyone else is told “This table supplies the row access grants for ‘region’, so only a project Architect can change its rows or delete it.”
Restriction Templates
Section titled “Restriction Templates”If you’ve built a restriction shape you want to reuse — the same value source, value type, and blank-value setting — save it once and apply it wherever it’s needed, rather than rebuilding it column by column.
- Save as Template…, from an open Row Access screen, saves the current column’s shape as a named, tenant-level template. It saves the shape only — no grants travel with it.
- Apply Template, on another project’s column, starts that column from the template’s shape, with every group’s grants empty for you to fill in.
Any project Architect can save and apply templates. Renaming (Rename…) and deleting (Delete) a template takes a tenant admin, since a template is shared across the tenant. Deleting one leaves the columns it was applied to restricted as they are.
Grants from SSO Group Attributes
Section titled “Grants from SSO Group Attributes”A column can also take its grants from your identity provider instead of from PlaidCloud. A project Architect opts a column in from Row Access by enabling Accept row-access grants from SSO group attributes.
Once enabled, a security group’s Keycloak attribute row_access.<column> sets the values that
group can see through that column — set the attribute on the group in Keycloak, and PlaidCloud
picks it up automatically for any group that already has access to the project. There’s nothing
further to configure in Row Access itself, and the grants stay in sync in both PlaidCloud’s own
queries and Superset dashboards.
Previewing Before You Save
Section titled “Previewing Before You Save”Row Access shows you what a change means before you commit to it:
- A Preview as footer shows how many of the table’s rows a group would see out of the total, with a sample of the rows themselves — pick any group in the grid to preview it. The preview refreshes after you save.
- Before you save, Row Access warns you about anything worth a second look: a group with nothing checked, and whether that means it sees all rows or none under your Everyone else setting; members with project access who aren’t covered by any group’s grant; or rows whose value in that column is blank (
NULL) that no group is granted.
What a Restricted Member Sees
Section titled “What a Restricted Member Sees”Wherever a restricted member reads a table by name, they get only their rows — Table Explorer’s grid, column facets and downloads, the table download dialog, apps and tools that read as the viewer, and dashboards. Table Explorer shows Rows limited by your access in its toolbar when the grid you’re looking at was filtered by your grants, and starting a download that will be filtered says “Rows limited by your access.” An Architect, who reads every row, sees neither notice.
Row Access and Queries You Write Yourself
Section titled “Row Access and Queries You Write Yourself”When Row Access limits your rows, you can run your own SQL queries on the project’s query screens and get exactly the rows your grants allow. You can export the results to CSV, Excel or JSON, and the export carries only those rows, with the Rows limited by your access notice. A project Architect reads every row, because an Architect can already grant themselves every value of every restricted column, and sees no notice.
A grant value that contains a percent sign (%) returns its granted rows on query screens, and downloading such a query as CSV works.
These are still declined for a member who is not a project Architect:
- Running your own free-form SQL on a Snowflake or Databricks project warehouse.
- Exporting the results of your own SQL from a Postgres-family warehouse.
- Exporting the results of your own SQL to another database.
- Typing a row condition into a fan-out step’s “Test row 1” preview.
- Reading a governed table as of an earlier snapshot (time travel).
Anyone declined is told which project it was, and pointed at the by-name readers: reading a table by name is the ordinary way apps and tools read data, and it applies the grants and returns the rows the viewer is entitled to. A column facet written as a plain query of a single table is read by name too, so it returns the member’s own values rather than being declined. A project Architect can run any of these as written.
Row Access filters a view on the governed column, so a view must keep that column under its original name. A view that drops or renames the governed column isn’t filtered.
Importing a spreadsheet is not, by itself, a query you wrote yourself — an ordinary file import reads no restricted data at all, it only creates the project’s own new table from the file you uploaded, so it is unaffected by Row Access. The exception is an import whose data projection is caller-supplied: a per-column expression, or a dynamic Excel mapping. On the interactive import screen those are declined the same way as the free-form-SQL surfaces above, and need a project Architect, because a caller-supplied projection could read another table. An import run from a workflow is unaffected either way, since it runs under the project’s own service identity.
Row Access and the Data Editor
Section titled “Row Access and the Data Editor”The Data Editor reads and writes a table directly, so Row Access applies to it on both sides.
Opening a restricted table in the Data Editor shows only the rows your security group is granted — the same filter that follows you into a dashboard, and it applies even when the restricted column is not one of the columns shown in the grid.
Saving your edits requires Architect on the project. A save replaces the table’s rows wholesale rather than reading them, and it is not filtered row by row, so editing data on a restricted table is treated as an authoring action rather than a read: anyone without Architect is told they need it. A table that carries no restricted column is edited as before, and so is any project that restricts no rows.
Loading Data Into a Restricted Table
Section titled “Loading Data Into a Restricted Table”When Row Access limits the rows you see, writing a DataFrame to a restricted table or clearing it with the Python client (PlaidConnection), or saving records to it or clearing it through the REST API, works on your own rows only. Replacing the table’s data replaces just the rows you can see, clearing it clears just those rows, and appending adds rows; every other row stays as it is.
A load carrying a row outside your access is refused before anything changes, and the error names each value outside your access with how many rows carry it. A row that leaves out the restricted column is checked as blank in that column, so it is refused unless your access includes blank rows. Replacing the data in a way that would change a column’s type is refused too; ask a project Architect to make that change.
Project Architects, members whose grants cover every row, and workflow runs, which run as the project’s service identity, write the whole table. A table with no restricted column is unaffected.
Anything else that replaces or removes a whole restricted table is refused while the table holds a row you cannot see, and nothing is changed: importing a file into it, copying a table into it from another project, switching or merging its branch, restoring it from an archive or as part of restoring the project’s held snapshots, deleting it, bringing it back after it was deleted, and any other action that empties the table before filling it again. A row whose restricted column is blank counts as one you cannot see unless your access includes blank rows. A table whose every row you can see, such as one you created holding only your own rows, can be replaced or deleted.
Each of these writes that a person makes, to any table, is recorded in the Security Log, including the ones that were refused. Workflow runs are not recorded there.
Loading a dimension from a table reads every row of the table, and everyone in the project sees the members it makes. So it runs only for someone who can read every row of the source table — a project Architect, a member whose grants cover every row, or a workflow run. A dimension that a restricted column reads still takes a project Architect, a tenant admin or a workflow run, as described under Who Can Change a Project That Restricts Rows. A table with no restricted column is unaffected. Anyone else is refused and told to ask a project Architect to load it, or to load it from a workflow.
Row Counts and Table Sizes
Section titled “Row Counts and Table Sizes”A table’s row count and storage size describe the whole table, not the rows you may see, so they would tell a restricted reader how many rows exist beyond their grant. On a project that restricts rows, anyone whose reads are filtered sees no row count or size for any of the project’s tables — in table lists, search, table details and snapshots, and through the AI assistant and AI tools connected over MCP. This covers every table in the project, including tables with no restricted column. The project’s own totals are hidden from them too — the total row count and size on Home, in project lists and search, and in project details.
The same applies to workflow run history. The rows read and written that each run and step recorded describe the whole table too, so for the same readers those figures are left blank — in a workflow’s and a step’s run history, and through the AI assistant and AI tools connected over MCP. For them, Weave does not check steps for rows or empty values unlike previous runs, since a finding would itself tell them the whole table’s count had moved.
Project Architects see the figures as before, and so does everyone on a project that restricts no rows.
Seeing What a Member Can See
Section titled “Seeing What a Member Can See”To check what a specific person can see, rather than what a group is granted: open Identity → Member, select the member, and open their Row Access tab. Its Effective Access column lists, for every restricted project and column, the values that member sees through all of their groups combined — including grants from SSO group attributes and blank (NULL) rows where a group’s grant includes them. A project Architect shows as Sees every row (Architect), since an Architect reads every row. It is the same answer PlaidCloud’s own queries apply, so it’s useful for support and for confirming a change did what you meant before someone reports a problem.
Clearer Errors
Section titled “Clearer Errors”Where Row Access blocks an action — a query you wrote yourself on a restricted project, an edit without Architect, a change to automation, a dimension or a grants source table, and the other cases described above — the message names what is restricted and what to do next, such as the Architects to ask, instead of a generic failure.
Row Access Is Audited
Section titled “Row Access Is Audited”Restricting or un-restricting a column, and editing a group’s grants, all write an entry to that project’s Project Log — who made the change and when. Open Analyze → Projects → Log on the project to see it. Nothing about Row Access has to be turned on to get this trail; it’s recorded the same way every other tracked change to a project is.
Setting Default Security Groups
Section titled “Setting Default Security Groups”To reduce the time needed for adding new members, identify a set of default security groups. This provides a baseline set of security groups for new members without needing to manually assign each person. The setting is available when adding a new security group if you check the box at the bottom of the Security Group window that reads “Assign to New Users by Default”.
Performing a Security Audit
Section titled “Performing a Security Audit”The security audit capability provides the ability to see group membership across all members and groups.
To perform a security audit:
- Open Identity
- Select Group audit under Security in the rail
As all tables in PlaidCloud are exportable as a CSV file format, the group member associations are reviewable outside of PlaidCloud for either historical purposes or just some fun off-line viewing.
To export from the “Security Group Audit” form:
- Open Identity
- Select Group audit under Security in the rail
- Click the small icon to the far right of “Username” in the table
- Click “Export CSV” or “Export XLXS” depending on your preference
Viewing Available Permission Settings
Section titled “Viewing Available Permission Settings”Each application being used in the workspace has specific available permissions. The security group permissions are based on these application permissions.
The complete list of available permission for each application is viewable from the Security Bin.
To access the Security Bin:
- Open Identity
- Select the “Security”
- Click “Security Bins” in the dropdown menu
To view the detailed security settings for each application, select the tags icon on the far left.
This available security settings information is informational only. For details on managing permissions, refer to the Managing Security Groups section above.